Load avarage spikes up to 20 (+)

[s.vasilyev]

Good day,

My zimbra server has crazy spikes on load avarage, this is what happens:
works as usual load is 0.2-0.5 then suddenly goes up to 20-40 load avarage for like few minutes then go normal again.

Load Avarage:


Now I installed htop, to monitor what exactly is going on during those spikes, this is what i get:

http://img833.imageshack.us/img833/542/highload1.png
http://img688.imageshack.us/img688/9826/highload2.png
http://img695.imageshack.us/img695/6475/highload3.png
http://img146.imageshack.us/img146/8176/highload4.png

I have 500+ users, only imap+web. Limit for imap connections is 2000.
12GB ram, Intel(R) Xeon(R) CPU E5620 @ 2.40GHz (16 cores)
Running on Ubuntu 8.04 64 Bit.
Zimbra version Release 6.0.8_GA_2661.UBUNTU8_64 UBUNTU8_64 FOSS edition.

I've done the follwoing fix High CPU spikes every 1 minute but still. got spikes, not so often tho.

Can any1 tell where to look for problem?

Thank you.

[s.vasilyev]

Please help?

[dalmate]

Can you know which process made that thing?
please run "top" command and see the output.

[hescominsoon]

Originally Posted by s.vasilyev

Good day,

My zimbra server has crazy spikes on load avarage, this is what happens:
works as usual load is 0.2-0.5 then suddenly goes up to 20-40 load avarage for like few minutes then go normal again.

Load Avarage:


Now I installed htop, to monitor what exactly is going on during those spikes, this is what i get:

http://img833.imageshack.us/img833/542/highload1.png
http://img688.imageshack.us/img688/9826/highload2.png
http://img695.imageshack.us/img695/6475/highload3.png
http://img146.imageshack.us/img146/8176/highload4.png

I have 500+ users, only imap+web. Limit for imap connections is 2000.
12GB ram, Intel(R) Xeon(R) CPU E5620 @ 2.40GHz (16 cores)
Running on Ubuntu 8.04 64 Bit.
Zimbra version Release 6.0.8_GA_2661.UBUNTU8_64 UBUNTU8_64 FOSS edition.

I've done the follwoing fix High CPU spikes every 1 minute but still. got spikes, not so often tho.

Can any1 tell where to look for problem?

Thank you.

What kind of hdd's?

[s.vasilyev]

Originally Posted by dalmate

Can you know which process made that thing?
please run "top" command and see the output.

If you click on images, it does show processes which taking 100% CPU time

Originally Posted by hescominsoon

What kind of hdd's?

Got software raids configured:
for operation system (Ubuntu 8.0.4 x64) got RAID1(2x250GB) as /dev/md0 ~ 220GB
for swap got RAID1 as /dev/md1 ~ 12GB (same as RAM)
for /opt/ directory got RAID10 (4x500GB) ~ 925GB

as of brand i think its Western Digital, but i really dont think problem is in them.

[hescominsoon]

with how many users you have you really should be running hardware raid. It's quite a bit faster as long as you get a good card. There's way more crappy ones out there than good ones. Also those 100% cpu lines could be stuck in i/o wait which that screen doesn't show. run top..not htop and when things are at 100% see if the 100% cpu corresponds with 100% or more in the WA colum. if it does..it's your disk setup. Are you running 7200, 10k, 15k disks?

[hescominsoon]

on top once you launch hit s then 1 to make the display every second then 1 again to break out each individual cpu. Take some screenies of it when the load is high..

[s.vasilyev]

Thank you i'll be looking forward for next spike

[s.vasilyev]

Ok i got it, CPU Load was over 20, but thats what i got, i dont think its HDD's

[hescominsoon]

no i/o wait..but there could be a problem with i/o anyway. Have you considered going up to the latest 10.04 LTS on your ubuntu? I'm not a huge fan of Ubuntu but i've seen issues with the base OS(Linux and Windows) cause weird issues like this.